Ethnolinguistic Favoritism in African Politics ONLINE APPENDIX Andrew Dickensy For publication in the American Economic Journal: Applied Economics yBrock University, Department of Economics, 1812 Sir Issac Brock Way, L2S 3A2, St. Catharines, ON, Canada (email:
[email protected]). 1 A Data Descriptions, Sources and Summary Statistics A.1 Regional-Level Data Description and Sources Country-language groups: Geo-referenced country-language group data comes from the World Language Mapping System (WLMS). These data map information from each language in the Ethnologue to the corresponding polygon. When calculating averages within these language group polygons, I use the Africa Albers Equal Area Conic projection. Source: http://www.worldgeodatasets.com/language/ Linguistic similarity: I construct two measures of linguistic similarity: lexicostatistical similarity from the Automatic Similarity Judgement Program (ASJP), and cladistic similar- ity using Ethnologue data from the WLMS. I use these to measure the similarity between each language group and the ethnolinguistic identity of that country's national leader. I discuss how I assign a leader's ethnolinguistic identity in Section 1 of the paper. Source: http://asjp.clld.org and http://www.worldgeodatasets.com/language/ Night lights: Night light intensity comes from the Defense Meteorological Satellite Program (DMSP). My measure of night lights is calculated by averaging across pixels that fall within each WLMS country-language group polygon for each year the night light data is available (1992-2013). To minimize area distortions I use the Africa Albers Equal Area Conic pro- jection. In some years data is available for two separate satellites, and in all such cases the correlation between the two is greater than 99% in my sample.